Indicating Lung Maturity of Preterm Births on LUS Images using the GLCM Statistical Properties --- A Pilot Study

The lung immaturity of premature infants can cause a composite of respiratory distress or even death. Lung ultrasound (LUS) is a non-radiation and real time technique, which has been applied to diagnose lung diseases. And the gray-level co-occurrence matrix (GLCM) can describe texture features of an image. The objectives of this study are to analyze texture features of LUS images using the GLCM properties and to investigate the relationship between GLCM statistical properties and the extent of lung maturity. For data processing, the region of interest (ROI) was selected and co-occurrence matrix of ROI was generated with predetermined offset parameters. Using the matrix, the values of statistical properties such as Contrast, Correlation and other four properties were computed, and the related mathematical statistics including mean and standard deviation were also calculated and investigated. Of all co-occurrence matrix properties, Homogeneity, Entropy, and Contrast presented higher correlation to the corrected gestational age. While Homogeneity and Entropy showed more reliable evaluation results with lower coefficient of variation. In conclusion, it demonstrated that co-occurrence matrix properties had the potential of evaluating the lung maturity of premature infants based on LUS, and Homogeneity and Entropy could provide more reliable and robust assessment results.